Closures of Relations

1
 Relational closures
• Three types we will study

– Reflexive
• Easy

– Symmetric
• Easy

– Transitive
• Hard
2
 Reflexive closure
• Consider a relation R:
(Note that it is not reflexive)

• We want to add edges to

make the relation reflexive

• By adding those edges,

we have made a non-
reflexive relation R into
a reflexive relation

• This new relation is called the reflexive closure of R

3
 Reflexive closure
• In order to find the reflexive closure of a
relation R, we add a loop at each node
that does not have one

• The reflexive closure of R is R U

– Where  = { (a,a) | a  R }
• Called the “diagonal relation”
– With matrices, we set the diagonal to all 1’s

4
 Example:
• Let R be a relation on the set { 0, 1, 2, 3 } containing the
ordered pairs (0,1), (1,1), (1,2), (2,0), (2,2), and (3,0)
• What is the reflexive closure of R?
• We add all pairs of edges (a,a) that do not already exist

0 1
We add edges:
(0,0), (3,3)

3 2

5
 Symmetric closure
• Consider a relation R:
(Note that it is not symmetric)

• We want to add edges to

make the relation symmetric

• By adding those edges,

we have made a non-
symmetric relation R into
a symmetric relation

• This new relation is called the symmetric closure of R

6
 Symmetric closure
• In order to find the symmetric closure of a
relation R, we add an edge from a to b,
where there is already an edge from b to a

• The symmetric closure of R is R U R-1

– If R = { (a,b) | … }
– Then R-1 = { (b,a) | (a,b) in R }

7
 Example:
• Let R be a relation on the set { 0, 1, 2, 3 } containing the
ordered pairs (0,1), (1,1), (1,2), (2,0), (2,2), and (3,0)
• What is the symmetric closure of R?
• We add all pairs of edges (a,b) where (b,a) exists
– We make all “single” edges into anti-parallel pairs
0 1

We add edges:
(0,2), (0,3)
(1,0), (2,1)
3 2

8
 Paths in directed graphs
• A path is a sequences of connected edges from vertex a
to vertex b

• No path exists from Start (a)

the noted start location

End (b)
• A path that starts and
ends at the same
vertex is called a Start (a)
circuit or cycle End (b)
– Must have length ≥1

9
Start (a)
 More on paths…
• The length of a path is the number of
edges in the path, not the number of
nodes

10
 Shortest paths
• What is really needed in most applications
is finding the shortest path between two
vertices

11
Transitive closure
The transitive closure would
contain edges between all nodes
reachable by a path of any length

12
 Transitive closure
• Informal definition: If there is a path from a to b, then
there should be an edge from a to b in the transitive
closure
• First take of a definition:
– In order to find the transitive closure of a relation R, we add an
edge from a to c, when there are edges from a to b and b to c
• But there is a path from 1 to 4 with no edge!
(1,2) & (2,3) (1,3)
R = { (1,2), (2,3), (3,4) } 4 (2,3) & (3,4) (2,4)
4

1 1
2 3 2 3
13
 Transitive closure
• Informal definition: If there is a path from a to b, then
there should be an edge from a to b in the transitive
closure

• Second take of a definition:

– In order to find the transitive closure of a relation R, we add an
edge from a to c, when there are edges from a to b and b to c
– Repeat this step until no new edges are added to the relation
• We will study different algorithms for determining the
transitive closure
4
• red means added on
the first repeat 1
• teal means added on 2 3
the second repeat 14
 Connectivity relation
• R contains edges between all the nodes reachable via 1 edge
• R◦R = R2 contains edges between nodes that are reachable via 2
edges in R
• R2◦R = R3 contains edges between nodes that are reachable via 3
edges in R
• Rn = contains edges between nodes that are reachable via n edges
in R

• R* contains edges between nodes that are reachable via any

number of edges (i.e. via any path) in R
– Rephrased: R* contains all the edges between nodes a and b when is a
path of length at least 1 between a and b in R

• R* is the transitive closure of R

– The definition of a transitive closure is that there are edges between any
nodes (a,b) that contain a path between them
15
 How long are the paths in a
transitive closure?
• Let R be a relation on set A, and let A be a set with n
elements. Consider a graph G with n nodes and some number of
edges
• Theorem: If there is a path (of length at least 1)
from a to b in R, then there is a path between a
and b of length not exceeding n
• Proof preparation:
– Suppose there is a path from a to b in R
– Let the length of that path be m
– Let the path be edges (x0, x1), (x1, x2), …, (xm-1, xm)
– That’s nodes x0, x1, x2, …, xm-1, xm
– If a node exists twice in our path, then it’s not a shortest path
• As we made no progress in our path between the two occurrences
of the repeated node
– Thus, each node may exist at most once in the path
16
 How long are the paths in a
transitive closure?
• Proof by contradiction:
– Assume there are more than n nodes in the path
• Thus, m > n
• Let m = n+1
– By the pigeonhole principle, there are n+1 nodes in the path
(pigeons) and they have to fit into the n nodes in the graph
(pigeonholes)
– Thus, there must be at least one pigeonhole that has at least two
pigeons
– Rephrased: there must be at least one node in the graph that
has two occurrences in the nodes of the path
• Not possible, as the path would not be the shortest path
– Thus, it cannot be the case that m > n

• If there exists a path from a to b, then there is a path

from a to b of at most length n
17
 Finding the transitive closure
• Let MR be the zero-one matrix of the relation R
on a set with n elements. Then the zero-one
matrix of the transitive closure R* is:

M R*  M R  M [R2]  M [R3]    M [Rn ]

Nodes reachable Nodes reachable Nodes reachable

with one application with two applications with n applications
of the relation of the relation of the relation

18
 Example:
• Find the zero-one matrix of the transitive
closure of the relation R given by:
1 0 1 1 2 1 2
M R  0 1 0
1 1 0 M [R2 ] 

3 3
M R*  M R  M [ 2]
R M [ 3]
R

1 0 1 1 0 1  1 1 1 
M [R2]  M R  M R  0 1 0  0 1 0   0 1 0 
   
1 1 0 1 1 0 1 1 1
19
 1 1 1  1 0 1 1 1 1
M [R3]  M [R2 ]  M R  0 1 0   0 1 0   0 1 0 
   
1 1 1 1 1 0 1 1 1
1 2

1 0 1 1 1 1 1 1 1 1 1 1
M R*  0 1 0  0 1 0  0 1 0  0 1 0
1 1 0 1 1 1 1 1 1 1 1 1 20
 Transitive closure algorithm
• What we did (or rather, could have done):
– Compute the next matrix M[iR], where 1 ≤ i ≤ n
– Do a Boolean join with the previously
computed matrix
• For example:
– Compute M[R2]  M R  M R
– Join that with M R to yield M R  M[R2 ]
– Compute M R  M R  M R
[ 3] [ 2]

– Join that with M R  M[R2 ] from above

21
 Transitive closure algorithm
procedure transitive_closure (MR: zero-one nn matrix)
A := MR
B := A
for i := 2 to n
begin
A := A  MR
B := B  A
end { B is the zero-one matrix for R* }

What is the time complexity? O(n4)

22
 Warshall Algorithm
Procedure Warshall (MR : rank-n 0-1 matrix)
W := MR
for k := 1 to n
for i := 1 to n
for j := 1 to n
wij := wij  (wik  wkj)
return W {this represents R*}
wij = 1 means there is a path from i to j going only through nodes ≤k.
Indices i and j may have index higher than k.

Time Complexity: O(n3)

23
Lemma:

24
Example: R={(1, 2), (1, 3), (2, 3), (3, 4) }.

Using Warshall’s algorithm find the transitive closure.

25
26
Example:

27